一种三维激光点云中建筑物立面渐进分割方法  

Gradual segmentation method for building facade in 3D laser point cloud

作　　者：卢秀山 邢恺强 刘如飞[2] 王鹏[2] LU Xiushan;XING Kaiqiang;LIU Rufei;WANG Peng(Institute of Ocean Engineering,Shandong University of Science and Technology,Qingdao,Shandong 266590,China;College of Geomatics,Shandong University of Science and Technology,Qingdao,Shandong 266590,China)

机构地区：[1]山东科技大学海洋工程研究院,山东青岛266590 [2]山东科技大学测绘科学与工程学院,山东青岛266590

出　　处：《测绘科学》2019年第12期7-13,共7页Science of Surveying and Mapping

基　　金：国家重大仪器设备开发专项（2013YQ120343）;国家重点研发计划项目（2016YFB0501705）;国家基础测绘科技计划项目（2016KJ0101）;山东科技大学人才引进科研启动基金项目（2016RCJJ004）;山东科技大学研究生科技创新项目（SDKDYC180311）

摘　　要：针对建筑物立面分割的问题,该文提出了一种三维激光点云中建筑物立面渐进分割方法。建立三维格网索引,通过分析建筑物立面在三维格网内的空间分布特征和二维平面格网内投影的线性分布特征,确定立面种子格网和投影线的拟合点,拟合投影线并基于种子格网约束生长完成每层格网中立面粗分割;使用RANSAC算法对粗分割后的立面点云进行面拟合,实现精细分割,并将各层格网中的立面进行合并,实现建筑物立面的完整分割。实验结果表明,该方法能有效地实现建筑物立面的精细化分割,有助于后续的建筑物精细三维模型重建。Aiming at the facade segmentation of buildings,agradual segmentation method for building facades in 3Dlaser point cloud is proposed in this paper.A three-dimensional grid index is established for the original point cloud,the facade seed grid and the fitting points of the projection line are determined by analyzing the spatial distribution characteristics of building facades in three-dimensional grids and the linear distribution characteristics of projections in two-dimensional plane grids,the projection line is fitted and the rough segmentation of the neutral plane of every layer is completed based on the seed grid constraint growth;The RANSAC algorithm is used to face-fit the rough-divided facades point cloud to achieve fine segmentation,and the facades in each layer of the grid are combined to completely segment the building facade.The results of experiments show that the method can effectively achieve the fine segmentation of the building facade and may contribute to the subsequent reconstruction of the fine 3D model of the building.

关 键 词：三维激光点云 建筑物立面分割 方向约束 RANSAC 

分 类 号：P237[天文地球—摄影测量与遥感]